Link Test and Flood Test

Link Test and Flood Test

In the PMP 450/450i/450m AP, the link/flood test option is used to evaluate the maximum capacity of an individual link or of the whole sector, and it is found under Tools -> Link Capacity Test.

The Link Test Mode supports the following options:

In the link tests with and without bridging, the link test task in software has higher priority than the bridge, which means that the traffic coming into the bridge will not be transferred to the MAC queues. Only the link test traffic to the designated VC (Virtual Circuit) is added to the MAC queue, and it is the only traffic going over the RF link. Note that if there was traffic in the MAC queue prior to the start of the link test, that data will be scheduled first. After that data is transferred, only link test data is sent over the link for the duration of the link test. 

In the Link Test with Bridging and MIR, once the MIR (Maximum Information Rate) is met for the VC, user traffic is also sent over the link.

In the Extrapolated Link Test, 64 packets are transmitted to the VC, and the corresponding performance extrapolated to determine the expected throughput that would have been achieved if the queue were completely full. The transmission of these packets does not interrupt the transmission of user traffic.

The four link test options behave the same way in PMP 450, PMP 450i and in PMP 450m.

Important configuration parameters for the flood test are:

  • Link Test VC priority: Options are High and Low Priority VCs or Low Priority VC only
    Note that if the High and Low Priority VCs option is selected, all queues all filled, but the high priority VCs will be scheduled first. Because those queues are filled, the scheduler never reaches the queues of the low priority VCs, unless CIR is configured for the low priority VCs. Also, high priority VCs are not scheduled in MU-MIMO mode, which means that no grouping is possible for these VCs.
    To test the MU-MIMO functionality, select Low Priority VC only.
  • MU-MIMO: Options are Enabled or Disabled. If MU-MIMO is disabled, the AP communicates to each VC in sector mode with no grouping; if MU-MIMO is enabled, the AP will attempt to group the VCs according to their spatial frequency.
  • Display results for untested VCs: Options are Enabled or Disabled
    In case the flood test involves a subset of the VCs registered in the sector, enabling this option adds the display of the traffic directed to the VCs that are not involved in the flood test.

Comparison between Link Test and Flood test to one VC

The following tests have the goal to show the difference between a Link Test and a Flood test to one VC.

First, a Link Test runs to one of the VCs registered in the sector, when there is no user traffic in the sector.

From the Link Test Mode drop down menu, the Link Test without Bridging option is selected, as the flood test also is without bridging. The results section for the Link test shows a DL throughput of 93 Mbps. This is the DL throughput achievable at 8x modulation with the selected frame configuration.

Next, the flood test is run, selecting as subset of VCs only the one VC that was used for the link test above.

 

As there is no user traffic in the link, and only one VC is part of the flood test, all groups have a size of 1, and the VC throughput is around 94 Mbps.

Without traffic, the link test and the flood test to one VC give the same result.

Let’s now add user traffic in the amount of 10 Mbps to five of the seven LP VCs in the sector. The link test is run first, and then the flood test, to one of the five VCs with user traffic.

The main difference between the link test and the flood test is that the link test stops the user traffic, and fills the queues with test data; on the other hand, the flood test transfers the user data first, and fills only the empty portion of the queues with test data.

The link test result shows again that the VC can receive 93 Mbps in the DL, the same as in the case with no traffic. The AP stops all five streams of data to the five VCs, and fills the queue of the VC it is running the link test with.

The flood test results show that the four VCs with user traffic that are not used in the flood test receive the expected 10 Mbps, and only the VC used in the flood test uses the remaining resources. This VC has a throughput of 10 Mbps of user traffic plus 61 Mbps of flood traffic, for a total of 71 Mbps.

Note the difference between the two tests. The link test shows the maximum throughput the VC can achieve in SU-MIMO mode (93 Mbps), but in order to show this capability the user traffic is interrupted. The flood test also shows the maximum throughput the VC can achieve in MU-MIMO mode (71 Mbps), but in order to shows this capability user traffic is not stopped, and the other four VCs receive the expected 10 Mbps. Additionally, the VC in the flood test also receives its 10 Mbps of user traffic during the test, while during the link test the VC does not receive its own user data.

 

For more details on how the flood test works, many examples of flood tests with and without user traffic, and with and without CIR configured, see the attached paper.

5 Likes

In the context of the article segment shown below would anyone be able to confirm or correct the definition of the terms highlighted?
Q1 MAC Queue
Seems like the queue that data would be placed on to await scheduling and transmission on the link in question?
Q2 I note reference to plural MAC queues. So one MAC queue for each link?
Q3 Bridge
Software component handling L2 traffic?
So traffic “coming into the bridge” would be traffic arriving at the AP for example from the local Ethernet port destined for transmission out a different link?
Q4 Link tests with/without Bridging
These 2 tests are intrusive correct? In other words the stop production traffic while the test is run since “only link test data is sent over the link for the duration of the link test”

Much appreciate any responses

Yes, exactly. The Canopy MAC queue receives packets from the Ethernet and queues them up to go over the RF. (or vice versa)

Yes, each Data Channel (old term is VC = Virtual Circuit) has its MAC queue. Note a SM can be optionally configured for multiple Data Channels.

Yes, correct.

Yes, these test will mostly likely impact production traffic. Note it is possible for other SMs still to get a little bit of data through during the test, but during Link Tests (except Extrapolated) the AP’s Bridge stops adding other packets to the queues, but if there is already packets in queue, they will still get transmitted.

Also other SM’s uplink traffic gets scheduled as normal during Link Tests, so you would expect in a busy sector to see lower uplink throughput during busy hour, then say if you ran the Link Test when AP had zero uplink traffic.

Hope I answered all your questions, let me know if you have any more.

3 Likes

Great @Charlie thanks a lot very helpful :slight_smile:

1 Like